Trusted Uncertainty in Large Language Models: A Unified Framework for Confidence Calibration and Risk-Controlled Refusal

📅 2025-09-01
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
This work addresses the lack of controllable refusal capability in large language models (LLMs). We propose the first unified framework for uncertainty calibration and risk-controlled rejection tailored to API-based black-box LLMs—requiring no fine-tuning and offering distribution-free theoretical guarantees. Methodologically, it integrates heterogeneous uncertainty signals—including sequence likelihood, self-consistency dispersion, retrieval compatibility, and tool feedback—into a lightweight calibration via temperature scaling and adaptive scoring, then enforces principled rejection using conformal risk control under user-specified error budgets. Key innovations include fine-grained factual alignment and interpretable refusal. Experiments across short-form QA, code generation, and retrieval-augmented long-text generation demonstrate substantial improvements over entropy- and logit-threshold baselines: lower calibration error, superior area under the risk–coverage curve, and higher coverage at fixed risk levels.

Technology Category

Application Category

📝 Abstract
Deployed language models must decide not only what to answer but also when not to answer. We present UniCR, a unified framework that turns heterogeneous uncertainty evidence including sequence likelihoods, self-consistency dispersion, retrieval compatibility, and tool or verifier feedback into a calibrated probability of correctness and then enforces a user-specified error budget via principled refusal. UniCR learns a lightweight calibration head with temperature scaling and proper scoring, supports API-only models through black-box features, and offers distribution-free guarantees using conformal risk control. For long-form generation, we align confidence with semantic fidelity by supervising on atomic factuality scores derived from retrieved evidence, reducing confident hallucinations while preserving coverage. Experiments on short-form QA, code generation with execution tests, and retrieval-augmented long-form QA show consistent improvements in calibration metrics, lower area under the risk-coverage curve, and higher coverage at fixed risk compared to entropy or logit thresholds, post-hoc calibrators, and end-to-end selective baselines. Analyses reveal that evidence contradiction, semantic dispersion, and tool inconsistency are the dominant drivers of abstention, yielding informative user-facing refusal messages. The result is a portable recipe of evidence fusion to calibrated probability to risk-controlled decision that improves trustworthiness without fine-tuning the base model and remains valid under distribution shift.
Problem

Research questions and friction points this paper is trying to address.

Calibrate uncertainty evidence into correctness probability
Enforce user-specified error budget via principled refusal
Reduce confident hallucinations while preserving coverage
Innovation

Methods, ideas, or system contributions that make the work stand out.

Unified framework for uncertainty calibration and refusal
Lightweight calibration head with temperature scaling
Conformal risk control for distribution-free guarantees
🔎 Similar Papers
2024-06-12Neural Information Processing SystemsCitations: 27
💼 Related Jobs
No related jobs found.
M
Markus Oehri
University of Liechtenstein
G
Giulia Conti
University of the Republic of San Marino
K
Kaviraj Pather
University of Mauritius
A
Alexandre Rossi
International University of Monaco
L
Laia Serra
University of Andorra
A
Adrian Parody
University of Gibraltar
R
Rogvi Johannesen
University of the Faroe Islands
A
Aviaja Petersen
Ilisimatusarfik (University of Greenland)
A
Arben Krasniqi
University of Prizren "Ukshin Hoti"